This paper defines a land cover classification technique based on the annual NDVI cycle. A similarity measure based directly on the components of the Discrete Fourier Transform is used to determine a pixels class membership. This Fourier component similarity measure produces an objective, computationally inexpensive and rapid method of classification that is able to classify rangeland vegetation by dominant shrub type, and which performs favorably compared to previously published classification techniques. By also defining a Fourier component based coverage measure this technique provides an estimate of vegetation coverage.
Figure 3: Fourier component similarity measure (FCSM) classification. White pixels are unclassified. They include water bodies, areas with relatively short snow covered
periods and some pixels that differ substantially from all the reference pixels used.
This page is maintaind by Jason Evans |
Last updated 31st January 2013